Artists Score Early-Round Win Against Image Generators
Court denies motions by Stability AI, Midjourney to dismiss charges in lawsuit
The first flurry of lawsuits brought by artists and authors against AI companies took a spaghetti-at-the-wall approach to stating claims. They all claimed direct copyright infringement for allegedly copying their work in the process of training AI models. But they often threw in claims of inducing or otherwise contributing to copyright infringement by the models’ users, along with claims for trademark infringement, unfair competition, violations of plaintiffs rights of publicity and other non-copyright causes of action.
The courts hearing those lawsuits, in many cases, have whittled down the initial complaints substantially, paring away the secondary liability and non-copyright claims as not adequately presented, leaving only the claims of direct infringement during training to go forward. With some of those dismissals, however, the courts granted plaintiffs leave to correct the deficiencies in their initial complaints and refile the charges.
At least one group of artists who had seen most of their charges against Stability AI, Midjourney and Deviant Art dismissed by U.S. District Court Judge William Orrick last year did just that. And this time, Judge Orrick let stand their claims of inducing copyright infringement by users and of false endorsement and trade dress infringement under the Lanham trademark act, at least for now (the Deviant Art dismissal was upheld but Runway was added as a defendant).
His 33-page ruling is potentially an important victory for copyright owners in their ongoing battle with AI companies. For the first time, some of their broader theories of harm will at least get a hearing and be subject to discovery.
Most worrisome for AI companies is the court’s analysis of the plaintiffs’ theories of direct and induced copyright infringement. Orrick distinguishes the inducement claim made against Stability AI and the Stable Diffusion foundation model, from earlier cases where a technology has been accused of inducing copyright infringement, such as in the “Sony Betamax” case in 1984.
The theory of this case is not similar to – for example – a case asserting contributory infringement based on the sale of VCRs where, after discovery, plaintiff had no evidence of defendant’s intent to induce infringement. The Supreme Court explained that, in those circumstances, intent could not be “based on presuming or imputing intent to cause infringement solely from the design or distribution of a product capable of substantial lawful use, which the distributor knows is in fact used for infringement” [citation omitted]. Instead, this is a case where plaintiffs allege that Stable Diffusion is built to a significant extent on copyrighted works and that the way the product operates necessarily invokes copies or protected elements of those works. The plausible inferences at this juncture are that Stable Diffusion by operation [sic] by end users creates copyright infringement and was created to facilitate that infringement by design.
Stability’s motion to dismiss the claim was denied.
Orrick also lets stand for now that Stable Diffusion may contain “compressed copies” of the artist plaintiffs’ works, and that, by distributing a version of the model, Runway directly infringed their exclusive rights of distribution.
First, the “Model Theory” is based on the theory that the Stable Diffusion1.5 product itself – after it was trained – is “an infringing Statutory Copy” of plaintiffs’ works or a “Statutory Derivative Work” because it represents a transformation of plaintiffs’ works. Second, Runway challenges plaintiffs’ “Distribution Theory” of infringement, based on allegations that Runway infringes plaintiffs’ exclusive distribution rights because distributing Stable Diffusion 1.5 is equivalent to distributing plaintiffs’ works. As with Stability, because Runway does not challenge the use of the images for training purposes, I need not address the other theories of direct infringement. However, I note that both the model theory and the distribution theory of direct infringement depend on whether plaintiffs’ protected works are contained, in some manner, in Stable Diffusion as distributed and operated. That these works may be contained in Stable Diffusion as algorithmic or mathematical representations – and are therefore fixed in a different medium than they may have originally been produced in – is not an impediment to the claim at this juncture.
Neither of those theories are close to being fully adjudicated, of course. This week’s ruling means only that the claims will not be dismissed outright prior to discovery. The judge could still dismiss them at the summary judgment phase after discovery is completed.
But it also means that the plaintiffs will at least get a look under the hood of the Stable Diffusion model and the business model behind its creation and distribution. What they find could eventually inform other pending cases against the defendants, or provide grounds for additional cases to be filed.
In another setback for AI companies, Judge Orrick also let stand for now the claim that Midjourney infringed plaintiffs’ “trade dress,” because it uses a “CLIP-guided model” trained on their artwork. The CLIP model was originally developed by OpenAI and combines images and natural language processing to create what the plaintiffs claim is a “trade-dress database” capable of recalling and recreating elements of each artist’s trade dress.
“Trade dress” is a subset of trademark and refers to the the overall look and design of a product that serves as a source identifier, such as the distinctive shape of a Coke bottle, and is protected under a combination of the federal Lanham Act and various state statutes and common law principles.
By allowing the claim to go forward for now, Judge Orrick, for the first time, is opening the door to non-copyright claims against the design and operation of generative AI models.